Repository navigation
Large mismatch for wall temperature with turbulent Prandtl model and y+ > 1 #2686
Description
Activity
- changed the title
[-]isothermal wall temperature is wrong with turbulent prandtl model[/-][+]Large mismatch for wall temperature with turbulent Prandtl model and y+ > 1[/+]on Jan 8, 2026 This is due to coarse wall meshes and treatment of temperature boundary condition as a flux. The computation of the heat flux using Fourier's law is inaccurate in turbulent flows when the mesh is coarse and no wall model is used. The only solution is to use a very fine near-wall mesh to properly resolve the turbulent boundary layer, or to have wall functions to compute a heat flux.
y+ = 5 leads to a discrepancy of 1% and then it quickly goes south from there.
Have a look at my comment in @Cristopher-Morales PR for the strong BC, I think @NAnand-TUD also mentioned this issue in an dev meeting.
Thanks for the detailed explanation. This makes sense — the mismatch seems tied to coarse near-wall resolution and treating the temperature BC as a flux without a wall model. The note about y+ ≈ 5 already giving ~1% discrepancy is especially helpful.
It sounds like the realistic paths forward are either enforcing a much finer near-wall mesh (y+ ≈ 1) or introducing a proper thermal wall function / strong BC. Happy to help test or prototype either approach if that’s useful.
Forcing temperature is probably better for convergence, the right BC is immediately applied. For the wall model treatment, we first need a more enhanced wall model, maybe based on y* instead of y+, so we can properly deal with adverse pressure gradients and stagnation regions.
@Nesar976 We had similar observations. @sanjayv99 might be able to join forces to enable this fix.
Thanks for confirming — it’s helpful to know this behavior has been observed independently.
I’d be interested in collaborating on this further, especially around testing or prototyping improvements to the wall treatment, and comparing notes on potential approaches. I’m keen to stay involved with this area of the codebase and contribute in a sustained way, including as part of longer-term efforts such as GSoC.
Happy to coordinate with you and @sanjayy99 on next steps.
when
TURBULENT_CONDUCTIVITY_MODEL= CONSTANT_PRANDTL_TURB
the temperature imposed at the wall is wrong and can be off by hundreds of degrees.
The case below has an inlet velocity of 40 m/s using INC_RANS and SST. The wall temperature is set to 1200K.
For a large part, the wall temperature is close to 400K.
The inlet fluid temperature is 300K. All fluid properties are constant.
This case is completely converged with residuals of -15